A Hierarchical Control Framework for AI-driven Interactive Art: Balancing Creator Intent, Algorithmic Autonomy, and Audience Agency

This paper investigates the multi-agent collaborative mechanisms underlying the creation of AI-driven interactive art. Drawing upon hierarchical control theory, it introduces the CAA (Creator–Algorithm–Audience) framework—a three-layered control architecture that deconstructs the roles and interrelations of creator intention, algorithmic autonomy, and audience agency within generative systems. Furthermore, the paper proposes a novel Creative Entropy (CE) evaluation model to quantitatively assess how generative freedom is distributed and negotiated among the three agents. Through empirical case studies of representative AI-based artworks, the study demonstrates how the CE model can measure structural openness, system complexity, and participatory engagement. By combining theoretical modeling and quantitative analysis, the findings contribute to the development of sustainable, interactive, and balanced co-creation paradigms in intelligent art systems.